Data labeling
LiDAR and 3D point cloud annotation for autonomy and robotics
3D cuboids, point-level semantic segmentation and camera–LiDAR fusion, labeled by annotators who qualify on gold-standard scenes before they touch your data.
- Pilot phase
- Two or more review layers
- You own every label
The hard frames are the ones that matter.
Perception teams rarely struggle with the easy scenes. They struggle with the pedestrian half-hidden behind a matatu, the sparse cluster at 80 metres, the motorbike weaving between lanes and the handcart no class list anticipated. That is where annotators disagree, and where inconsistent labels make a model hesitate.
We agree cuboid conventions up front, covering heading, truncation, occlusion levels and unusual road users, then track objects across sweeps so identities stay consistent through an entire sequence.
Used for
- Autonomous vehicle perception
- Warehouse and delivery robotics
- Drone and aerial survey data
- Construction and mapping
- Smart-infrastructure sensing
Industries: Autonomous vehicles · Manufacturing & robotics · Geospatial & satellite
What our LiDAR & 3d point cloud annotation covers
Combine any of these in one workflow and one guideline.
3D cuboids
Oriented boxes with heading, size and attributes, fitted tightly to the point cluster.
Point-cloud segmentation
Semantic and instance labels at the point level for drivable area, vegetation, structures and more.
Sensor fusion
Camera and LiDAR linked so every 3D object has a matching 2D label.
Multi-sweep tracking
Consistent object IDs through sequences of sweeps.
Lanes & road edges
3D polylines for lane markings, curbs and road boundaries.
Mapping features
Poles, signs and static infrastructure for HD-map datasets.
Formats and tools
Delivery formats
- KITTI
- nuScenes
- OpenLABEL
- PCD + JSON
- Custom schema
Tools we work in
- CVAT
- Segments.ai
- Encord
- Supervisely
- Your in-house tool
Pricing
3D cuboids from $0.12 per cuboid, and from $0.09 at volume. Point-level segmentation is quoted after a sample.
How we keep it accurate
Every project runs on the same four-check QA process. See how our QA works.
Scene-level QA
Reviewers check whole scenes, not single objects, so missed far-range clusters are caught.
Class-by-class agreement
Inter-annotator agreement and defect rates are tracked per class to surface problems early.
Lead-annotator escalation
Uncertain scenes are routed to a lead annotator for a final decision.
How a project runs
Scope and guidelines
We review a sample, agree classes, edge cases and acceptance criteria, and sign your NDA.
Pilot phase
Pilot phase helps you judge quality, speed and communication.
Certified team
Annotators pass a gold-set test before they are cleared for production.
Production and reporting
Batches with QA reports, in your format, on an agreed schedule.
LiDAR & 3D point cloud annotation questions
Do your annotators have 3D experience?
3D annotators need at least six months of 2D annotation experience and pass a task-specific 3D assessment before joining a project.
Can you work with our sensor setup?
Yes. Share your calibration files and coordinate conventions, and we configure the tool and guidelines around them.
How do you handle sparse or distant points?
Guidelines set minimum point counts and distance rules per class, with examples, so far-range objects are labeled consistently.
Send a sample. See the labeling quality first.
Share a few files and a line about your goals. A project lead replies within one business day with an approach, a timeline and a price.